Multiprocessor Architectures for Programmability

Dr. Ceze is from the University of Illinois at Urbana-Champaign
While multiprocessor hardware is finally becoming ubiquitous, enticing
many programmers to write parallel programs is going to be very
challenging. For this reason, I believe that the main problem that
confronts computer architects today is designing computer
architectures that help simplify parallel programming.

In this talk I will present two novel computer architecture techniques
that help simplify parallel programming. The first one is Bulk — a
framework of primitives for manipulating in hardware sets of addresses
in bulk. Bulk's primitives are used as building blocks to support
interactions between multiple threads, enabling high-programmability
environments such as thread-level speculation, transactional memory,
and high-performance sequential memory consistency. The second
technique is architectural support for data-centric
synchronization. The technique, called Colorama, associates
concurrency control constraints with data, providing an attractive
alternative to the traditional code-centric approach of locks and
transactions. Together, these two techniques offer promising
directions in the critical area of novel multiprocessor architectures
for programmability.
Luis Ceze is a PhD candidate in the Department of Computer Science at
the University of Illinois Urbana-Champaign (UIUC). His doctoral
research has focused on computer architectures to decrease the
hardware complexity and improve the programmability of multiprocessor
systems. He has co-authored over 20 papers in computer architecture,
programming models, and systems. He has also participated in the Blue
Gene, Cyclops, and PERCS projects at IBM. He has received awards from
UIUC for research and academic accomplishments, and from IBM for
contribution to the Blue Gene project.